mirror of
https://github.com/google-gemini/gemini-cli.git
synced 2026-05-15 06:12:50 -07:00
gemini-cli[bot]
08647b680a
I've investigated the triage failure and identified several issues:
1. **Failure Comment**: The automated triage was failing and posting a literal message containing `${process.env.RUN_URL}` because it used single quotes instead of backticks in the `github-script` step.
2. **Authentication**: The `Run Gemini Issue Analysis` step was explicitly setting `GITHUB_TOKEN: ''`, which appears to cause the `google-github-actions/run-gemini-cli` action to fail in some environments.
3. **Redundancy**: Both the automated and scheduled triage workflows were triggered on issue creation, leading to redundant analysis and noise.
4. **Noisy Comments**: The scheduled triage was posting a reasoning comment for every issue it triaged, which is often unnecessary for routine categorization.
### Changes:
- **Fixed `gemini-automated-issue-triage.yml`**: Removed the empty `GITHUB_TOKEN` and fixed the failure comment formatting so it correctly links to action logs.
- **Consolidated `gemini-scheduled-issue-triage.yml`**:
- Removed the redundant `issues: opened/reopened` triggers.
- Removed the empty `GITHUB_TOKEN`.
- Updated the analysis prompt to strongly discourage the use of example placeholders (like issue #123).
- Modified the label application script to only post comments when `status/need-information` is recommended.
- Added robust error handling to the label application loop so that a failure on one issue doesn't block the entire batch.
These changes should resolve the "broken triage" reports and reduce noise in the repository.